Taking the former Yugoslavia and some of its successor states as a case study, this article examines the concepts of ethnicity and mother tongue (with a side glance at religion) as employed in recent population censuses. A special focus is on the sometimes considerable discrepancies between the ethnic and linguistic affiliations of the respondents, a phenomenon commonly disregarded but potentially opening the door to manipulation of census returns in the service of particular interests. It is shown that, while having separate rubrics for the two features is in itself a useful methodological tool, their mutual relationship should be carefully investigated rather than simply taken for granted.